Full motion detection system with post-processing

Abstract

This paper presents a full motion detection system with post-processing applied to video surveillance. Motion detection is performed based on background subtraction (BGS). Our purpose is to show how an appropriate post-processing improves segmentation result provided by a BGS technique from the literature. First, BGS is performed using the codebook algorithm [8]. Post-processing is then applied on the raw segmented images to enhance the segmentation quality. The proposed post-processing proceeds in two steps. The first step consists in removing shadows and highlights based on color data. In the second step connected component labeling is used to extract pertinent foreground blobs. The full system is evaluated with video surveillance sequences under adverse illumination conditions.